Abstract
Recently, significance of human-machine interface by voice recognition techniques has been increased. Actually those techniques have been adapted to robots, ubiquitous system, cellular phone, auto-translation system and so on. Recognition researches of meaning ingredient included in human voice have been carried out mainly. However, there are many problems about the specification of speaker. The former recognition methods such as STFT, FFT, LPC and cepstrum analysis have many problems on recognizing accurately meaning, speaker characteristics and emotion state of speaker because human voice is a transient signal. Therefore, this study adopts the wavelet analysis, which has been using widely to analyze transient signals on Time-Frequency domain. This study was performed to extract the fundamental frequency using discrete wavelet analysis due to adopt a continuous wavelet analysis for the transition, it was revealed that a determination can be made between individuals.
